Epidemiology and Outcomes of Clostridium difficile-Associated Disease Among Patients on Prolonged Acute Mechanical Ventilation FREE TO VIEW

In the United States, patients requiring prolonged acute mechanical ventilation (PAMV [defined as mechanical ventilation [MV] for ≥ 96 h]) constitute one-third of all hospitalized patients undergoing MV. Despite representing a minority of all patients needing MV, those defined as needing PAMV consume two-thirds of the hospital resources devoted to the care of all MV patients.¹ Annually, hospital costs for PAMV exceed $16 billion, and the median hospital length of stay (LOS) among those patients requiring PAMV is 17 days.¹ The burden of PAMV will rise in the future because projections² suggest that the number of PAMV patients discharged from the hospital will double between 2000 and 2020, reaching > 600,000 cases annually. Because of their prolonged hospitalization and likely concomitant exposure to antibiotics, patients requiring PAMV are at risk for the development of superinfections with organisms such as methicillin-resistant Staphylococcus aureus and Clostridium difficile-associated disease (CDAD).

C difficile in particular has emerged as an important pathogen responsible for increasing morbidity and mortality.³ Since 2000, multiple hospital-based outbreaks of a new hypervirulent BI/NAP1/027 strain of C difficile have been reported^4,5 in the United States, Canada, and Europe. Advanced age, multiple comorbidities, and clear exposure to antibiotics are traditional risk factors for CDAD, although several reports^6,7 have identified community-acquired CDAD among populations of children, pregnant women, and other groups without these risk factors. Reflective of this changing epidemiology is the overall increase in the volume and incidence of hospitalizations with CDAD in the United States among both adults and infants.^3,8

Among critically ill patients, CDAD is responsible for a 6% incremental increase in the risk of death.⁹ Although the epidemiology and outcomes of CDAD in the resource-intensive PAMV population are not well understood, attention to the prevention of nosocomial complications in this population may improve not only clinical but also economic outcomes. We hypothesized that PAMV patients represent a unique cohort of subjects who are at increased risk for CDAD, and that CDAD in those patients with PAMV is associated with greater hospital mortality, LOS, and costs.

No human subjects were enrolled specifically in this study. This study was exempt from regulations guiding the protection of human subjects because it is a secondary analysis of a publicly available data set.

We utilized a single year (2005) of data from the Healthcare Costs and Utilization Project/Nationwide Inpatient Sample (NIS) of the Agency for Healthcare Research and Quality. The NIS consists of a stratified sample of hospital discharge records from approximately 1,000 participating facilities, representing about 20% of all community hospitals in the United States. The unit of reporting in this database is a hospital discharge. The NIS database includes data on patient demographics, diagnoses and procedures, in-hospital mortality, as well as hospital charges and LOS for each discharge. Additional data files, linkable to discharges in the NIS database, provide data on hospital characteristics, illness severity measures, and cost-to-charge conversion coefficients for each individual institution in the database. The Agency for Healthcare Research and Quality undertakes an assessment of completeness and data quality, and documentation is provided with the data set.¹⁰

We identified patients undergoing PAMV based on the presence of the International Classification of Diseases, ninth revision, clinical modification (ICD-9-CM) procedure code 96.72 in any discharge diagnosis field. The presence of CDAD was identified using the ICD-9-CM code 008.45 in any discharge diagnosis position.

The primary outcome of interest was hospital mortality. Secondary outcomes were hospital LOS (in days) and hospital costs (in US dollars).

We compared outcomes among all PAMV patients who were discharged from the hospital with a concomitant diagnosis of CDAD (CDAD+) to those PAMV patients who were discharged from the hospital without CDAD (CDAD−). We examined demographic, clinical, hospital, and discharge characteristics in these groups. Crude outcomes of CDAD+ hospitalizations were compared with CDAD−. Mean (SD) and median (interquartile range [IQR], 25 to 75%) values were calculated for continuous variables, and counts and proportions calculated for categorical variables. Continuous variables were compared between the two groups using the Student t test. Categorical variables were compared using the χ² test. All inferences were two tailed. Statistical significance was defined to be present at α = 0.05.

To adjust for confounding, we developed propensity score models based on patient demographics (age, race, and gender), patient type (urgent, emergent, elective, and trauma), hospital admission source, weekend hospital admission status, comorbidities (based on the Elixhauser classification¹¹), number of discharge diagnoses, and hospital characteristics (eg, hospital size and location) [Table 1]. Propensity score matching was conducted using a greedy 5:1 digit algorithm.^12,13 Stratification matching and kernel matching were also conducted to examine the consistency of the estimates via different matching methods.^12–14 We examined both spline terms and medically plausible interactions with age (one age interaction term was ultimately included). The model fit was assessed with the Hosmer-Lemeshow goodness-of-fit and the area under the receiver operating curve. Statistical analyses were performed using a statistical software package (Stata/SE, version 10.1; Stata Corp; College Station, TX).

Of the 64,910 discharges of patients with PAMV in 2005, CDAD was noted in 3,468 (5.3%). Among those patients with CDAD, this represented the principal diagnosis in 142 (4.1%). Table 1 shows the baseline characteristics of subjects undergoing PAMV. Compared with CDAD− patients, CDAD+ patients were older, and the prevalence of CDAD+ escalated with increasing age, stabilizing above the age of 70 years (Fig 1). CDAD+ discharges were less likely to be men or white. Although the urgency of hospital admission was similar in the two groups, those diagnosed with CDAD were less likely to have been initially admitted to the hospital following trauma (0.5% vs 1.2%, respectively). Those patients whose hospital admission was complicated by CDAD were 40% more likely to have been initially admitted from another health-care facility (relative risk, 1.39; 95% confidence interval [CI], 1.28 to 1.51). The burden of various comorbid illnesses was similar in both populations, and in both groups acute respiratory failure was the most prevalent principal diagnosis (Table 2).

Geographically, the burden of CDAD in PAMV patients was not distributed uniformly across the United States (Table 1). Despite the fact that the highest proportion of PAMV discharges occurred in the south (36.5%), more than one-third of cases of CDAD in this population arose in hospitals in the northeast. CDAD also varied as a function of hospital characteristics and was more common in urban teaching hospitals.

Table 3 presents the unadjusted primary outcomes. There was no difference in hospital mortality between CDAD+ and CDAD− patients discharged (32.6% vs 33.0%, respectively; p = 0.598). In contrast, the median hospital LOS was substantially greater among those patients with CDAD (25 days; IQR, 15 to 40 days) vs those without (17 days; IQR, 11 to 27 days). Reflecting this 8-day difference, the median costs were also higher among CDAD+ than CDAD− patients discharged ($57,607 [IQR, $34,272 to $96,936] vs $42,785 [IQR, $26,330 to $70,865], respectively). As for hospital discharge destination, those patients discharged to a skilled nursing facility were more likely to be in the CDAD+ than the CDAD− group (48.3% vs 38.6%, respectively), whereas routine discharges to home were more frequent in the CDAD− group than in the CDAD+ group (13.7% vs 7.0%, respectively; p < 0.001) [Table 3].

For the adjusted analysis, the propensity score was developed in 63,104 patients (2.8% of the discharge records had missing values). The model fit the data well (Hosmer-Lemeshow statistic, 8.75 [p = 0.36]; area under the receiver operating curve, 0.69 [95% CI, 0.68 to 0.70]) and was adequately balanced on all predictors. In the multivariable analysis adjusting for the propensity for CDAD, there was a small but statistically significant difference in mortality by CDAD status. In the matched analysis of 3,370 CDAD+ and 3,363 CDAD− patients, those with CDAD were somewhat less likely to die while hospitalized (adjusted relative risk, 0.89; 95% CI, 0.80 to 0.98). CDAD+ status was associated with an independent increase in the hospital LOS by 6.1 days (95% CI, 4.9 to 7.4 days) and an increase in total costs of $10,355 (95% CI, $7,540 to $13,170). The kernel and stratification matching methods produced similar estimates with overlapping CIs (data not shown).

Our analysis demonstrates that CDAD affects 1 in 20 hospitalized patients requiring ≥ 96 h of MV. This infection rate of 530 cases per 10,000 hospital admissions is strikingly high compared with that noted in the general hospitalized population (11.2 cases per 10,000 hospitalizations).³ Moreover, CDAD is associated with a 6-day prolongation of the hospital LOS and an additional hospital cost of > $10,000. Despite the overall high in-hospital mortality rate in PAMV patients, C difficile infection did not appear to worsen survival rates. Finally, a higher proportion of the CDAD+ group was discharged to a SNF, although more of them were admitted from a chronic care facility as well.

As we have reported previously, the PAMV population is large and growing. Representing approximately 300,000 discharges from US hospitals in 2000, this population is projected to nearly double by 2020.² Importantly, although the PAMV population comprises on-third of all MV patients in US institutions, it is responsible for two-thirds of the total annual expenditures in the overall MV population.¹ Taking into account their historical rate of growth, annual hospital bed occupancy and costs, as well as overall inflation, we have projected that the PAMV population may consume a total of 10.3 million hospital days and $64 billion in hospital costs in 2020 unless the processes of care change dramatically.¹⁵ As such, these facts combined with our current observations indicate that those needing PAMV represent an important population that is both at risk for and serving as a reservoir of C difficile.

A growing body of evidence^3,4,16–19 points to an increasing frequency and case fatality in CDAD hospitalizations in the United States. Clinical reports^16,17 from Quebec, Canada, and from Pittsburgh, PA, documented an approximate quadrupling of the incidence in CDAD cases earlier in this decade. This was followed by a report from Loo and coworkers,⁴ who in a surveillance effort in 12 hospitals in Quebec during the first half of 2004 noted an incidence of 22.5 cases per 1,000 hospital admissions of CDAD associated with increased mortality and morbidity. These major outbreaks were due to a new hypervirulent clostridial strain, BI/NAP1/027, whose unique features included increased fluoroquinolone resistance, the presence of a binary toxin, and a deletion within the tcdC gene responsible for the down-regulation of toxin production, resulting in 16-fold and 23-fold increases, respectively, in the production of toxins A and B.^5,18 By the end of 2007, 38 of the 50 US states had reported cases of C difficile infection with this deadly strain.²⁰ At the same time, reports^3,18,21 of national estimates have indicated increasing volumes, incidence, and case fatality of hospitalizations with CDAD.

Although past epidemiologic research dealing with CDAD suggests the increasing spread and virulence of this syndrome, few reports have explicitly addressed patients in the ICU, particularly those who, by virtue of a prolonged need for MV and ICU care, are at higher risk for exposure to both antibiotics and C difficile spores. Consequently, little is known about the epidemiology of CDAD or its effect on outcomes among ICU patients. Kenneally and coworkers⁹ found the overall unadjusted 30-day mortality rate in a cohort of 278 ICU patients with CDAD equaled 36.7%, along with a 6.1% CDAD-attributable mortality rate. Marra et al,²² in a single-center cohort study of 58 cases of CDAD among ICU patients, observed a hospital mortality rate of 27.6%. Our results confirm these observations and build on them. First, we note a crude mortality rate similar to what has been reported previously in critically ill subjects with CDAD. Second, and more importantly, through the use of a large sample derived nationally, we demonstrate that the burden of CDAD in the ICU represents an issue that is not geographically confined to selected areas of the country or selected types of institutions. As such, our observations suggest that urgent efforts are needed to address CDAD in this high-risk subgroup of ICU patients.

Our finding of a modest protective effect of CDAD in terms of hospital mortality is intriguing. First, it may be an artifact of the data representing an immortal time bias. That is, because it is not possible to determine the timing of the onset of CDAD relative to the entire hospital course, this ostensibly protective effect may reflect an imbalance in time at risk for developing CDAD; in other words, a patient in whom CDAD develops has to survive long enough to get it, thus eliminating those patients who are at a higher risk for an earlier death. Alternatively, the mortality imbalance may have arisen due to confounding variables we were unable to control for. Specifically, we could not measure severity of illness. The overall economic burden of CDAD, though, confirms the importance of this syndrome irrespective of its impact on mortality.

Our study has a number of strengths. Because we relied on a large nationally representative sample of PAMV discharges, our results are broadly generalizable to all US hospitals. Confirming this further is our finding^21,23 of geographic variability in the rates of CDAD, which mirror those reported previously among all US hospitalizations. Additionally, the sample size and number of covariates in the data set allowed us to adjust for confounding in a statistically rigorous manner. At the same time, there are a number of limitations. First, we relied on administrative codes, and not on clinical data, to identify both PAMV and CDAD, which may have predisposed our case definitions to error. However, methodologies similar to ours have previously been employed successfully in both PAMV and CDAD work.^21,24–27 In fact, a single-center study²⁵ indicated that the CDAD ICD-9-CM code is able to identify this condition with a high degree of accuracy. Second, because of the cross-sectional nature of the data set, it was not possible to ascertain what came first, PAMV or CDAD. We note, however, that the covariate adjustment was made with variables that were present at baseline and thus were not subject to time-dependent changes. Third, because we were unable to pinpoint the time of onset of CDAD, the disparities in the adjusted hospital LOS cannot be directly attributed to CDAD itself; that is, it is possible that at least some of the increase in the LOS (and its related costs) in the CDAD+ group was a risk factor for CDAD rather than a consequence of it. At the same time, at least some of the adjusted LOS difference is likely to be directly attributable to CDAD because nosocomial complications in general are known to affect LOS and costs. Finally, as in any cohort study, residual confounding is possible due to unmeasured variables not found in this database, although our propensity scores are based on a large number of covariates.

In summary, we have documented that CDAD is a frequent complication of PAMV, affecting 5.3% of this population. It is strongly associated with prolonged hospital LOS and an increase in hospital costs. Irrespective of the causation, the association we have detected is sufficient to indicate that these patients are an important reservoir for this disease, and, hence, should become a focus for efforts at prevention and control. Future studies are needed to quantify the resource utilization in this population that is directly attributable to CDAD, as well as to focus on effective prevention and treatment of this complication. Among other populations of hospitalized patients, CDAD epidemics can be mitigated with such measures as housekeeping interventions, handwashing, isolation and contact precautions, as well as with careful antibiotic stewardship.^4,28,29 Because these preventive measures are simple, sensible, and inexpensive, they should be implemented now in the ICU to avert the adverse outcomes associated with CDAD in critically ill patients. Given the resource-intensive nature of the PAMV population, as well as the economic burden of CDAD, clinical and cost effectiveness of aggressive measures aimed at preventing this complication need to be examined because such measures are likely to result not only in improvements in individual patient outcomes but also in increases in the efficiency of health-care delivery.

Author contributions: Dr. Zilberberg contributed to the conception, design, results interpretation, and drafting of the manuscript. Dr. Nathanson contributed to the development and execution of the analysis plan and the drafting of the manuscript. Mr. Sadigov contributed to the development and execution of the analysis plan and the drafting of the manuscript. Dr. Higgins contributed to the conception, results interpretation, and drafting of the manuscript. Dr. Kollef contributed to the conception, results interpretation, and drafting of the manuscript, and Dr. Shorr contributed to the conception, design, results interpretation, and drafting of the manuscript.

Financial/nonfinancial disclosures: Dr. Zilberberg has participated in a speakers' bureau supported by a grant from and has been a consultant to ViroPharma, the manufacturer of oral vancomycin. Drs. Nathanson, Higgins, Kollef, and Shorr, and Mr. Sadigov have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.